Researchers from Shanghai Children’s Hospital have developed and validated a web-based nomogram designed to predict the risk of subtherapeutic infliximab trough levels in children with Crohn’s disease. Published in Frontiers in Medicine, this retrospective analysis addresses a critical gap in pediatric gastroenterology, where therapeutic drug monitoring often lags behind adult practices due to logistical and financial barriers. The tool aims to shift clinical management from reactive monitoring to proactive, data-driven prediction, allowing clinicians to identify patients at risk of inadequate drug exposure before treatment failure occurs.
The study included 205 infliximab trough measurements from 74 pediatric patients diagnosed with Crohn’s disease who were receiving infliximab therapy at Shanghai Children’s Hospital. Infliximab, a monoclonal antibody targeting tumor necrosis factor-alpha, is a cornerstone treatment for inducing and maintaining remission in moderate-to-severe pediatric Crohn’s disease. However, a significant challenge in clinical practice is the loss of response to this therapy, which has been attributed to inadequate trough levels that fall below the therapeutic window.
Current pediatric guidelines recommend maintaining infliximab trough levels at or above 5 micrograms per milliliter to ensure efficacy. In this cohort, 108 of the 205 measurements, representing 52.7% of the observations, exhibited subtherapeutic levels.
Predictive features were selected using a combination of LASSO and Boruta regression algorithms to enhance stability and reduce selection bias in the relatively small sample size. The final model incorporated six independent predictors: dose escalation status, alkaline phosphatase, albumin, fecal calprotectin, erythrocyte sedimentation rate, and the Pediatric Crohn’s Disease Activity Index.
The nomogram demonstrated good discrimination in both the training and testing cohorts. The area under the receiver operating characteristic curve was 0.762 in the training set and 0.783 in the independent testing set. Bootstrap internal validation yielded a corrected AUC of 0.733 with an optimism value of 0.029, suggesting limited overfitting. Calibration curves indicated moderate agreement between predicted and observed probabilities, while Decision Curve Analysis revealed a potential net clinical benefit across a broad range of threshold probabilities, from 0.16 to 0.86 in the training data.
SHapley Additive exPlanations analysis provided interpretability for the model’s predictions. The analysis indicated that dose escalation and higher albumin levels were associated with protective effects against subtherapeutic trough levels. Conversely, higher levels of fecal calprotectin, erythrocyte sedimentation rate, alkaline phosphatase, and the Pediatric Crohn’s Disease Activity Index were identified as risk factors. The authors noted that dose escalation should be interpreted as a surrogate for previous disease severity or inadequate response rather than a direct causal pharmacokinetic effect, as it reflects prior clinical management decisions.
The model’s clinical utility was further evaluated through risk stratification using an optimal cutoff probability of 0.57, determined by maximizing Youden’s index. Patients classified as high-risk based on this threshold exhibited a significantly higher incidence of subtherapeutic trough levels compared to the low-risk group. In the training dataset, the incidence was 81.25% in the high-risk group versus 31.13% in the low-risk group. In the testing dataset, the corresponding figures were 92.31% and 54.55%, respectively. These findings suggest that the nomogram can effectively distinguish between patients with varying risks of inadequate drug exposure.
A complete-case analysis, which excluded participants with missing data, produced results consistent with the primary imputation analysis, with the AUC remaining stable at 0.793. Additionally, a sensitivity analysis using age- and sex-specific Z-scores for alkaline phosphatase confirmed that the association between this marker and subtherapeutic levels was robust to physiological variations associated with growth. This is particularly important in pediatric populations, where alkaline phosphatase levels naturally fluctuate due to increased osteoblastic activity.
Despite these promising results, the study has several limitations that highlight the need for further research. The retrospective, single-center design and small sample size limit the generalizability of the findings. The authors emphasize that external validation in larger, multicenter prospective cohorts is necessary to establish the model’s broader applicability. Furthermore, the model relies on routine clinical variables that reflect downstream systemic inflammation rather than upstream biological drivers, such as genetic polymorphisms or tissue-specific drug dynamics.
The development of this interactive web-based tool represents a step toward personalized medicine in pediatric Crohn’s disease. By transforming multiple predictors into an intuitive, quantifiable risk assessment, the nomogram provides clinicians with a practical tool for guiding infliximab therapy. However, the authors caution that the tool is intended to assist clinical judgment rather than replace it. The 0.57 cutoff has not yet been prospectively validated in real-world clinical decision-making, and further work is required to translate these findings into actionable decision rules.
As the incidence of pediatric Crohn’s disease continues to rise globally, the need for effective and accessible monitoring tools becomes increasingly urgent. Subtherapeutic infliximab levels not only lead to inadequate treatment response but may also impair growth and quality of life in children. By enabling early identification of patients at risk, this nomogram may facilitate closer pharmacokinetic monitoring and support timely dose optimization. The study underscores the importance of developing predictive models tailored to pediatric populations, as drug clearance in children differs significantly from that in adults.
In conclusion, this study successfully developed and validated a web-based nomogram for predicting the risk of subtherapeutic infliximab trough levels in pediatric Crohn’s disease. The model achieved stable discriminative performance with limited overfitting, offering a potential means to enhance the precision and data-driven management of infliximab therapy. While the tool shows promise, it remains a preliminary prediction instrument that requires further validation in diverse clinical settings to fully establish its clinical utility and generalizability.
Measurements of infliximab trough levels and anti-infliximab antibodies were carried out by Suzhou Herui Aibidi Medical Laboratory using a time-resolved fluorescence immunochromatographic assay. Operating with excitation and emission wavelengths of 365 nanometers and 610 nanometers, this assay provided a linear detection range from 0.4 to 20 micrograms per milliliter, maintaining both intra-assay and inter-assay coefficients of variation below 15 percent.
The dataset captured longitudinal practice patterns, with 68 of the 74 participants contributing between two and five separate trough assessments across treatment courses. The patient-level split allocated 60 unique children representing 170 measurements to the training cohort and 14 children representing 35 measurements to the testing cohort. Fecal calprotectin was missing in 16.6 percent of observations and tumor necrosis factor-alpha in 11 percent, while all other candidate parameters showed missing rates below 5 percent. Notably, baseline characteristics revealed an uneven sex distribution between the partitions, as male patients accounted for 75.9 percent of the training cohort but only 22.9 percent of the testing cohort.
With 85 positive outcome events in the training group across six finalized predictors, the model attained an events-per-variable ratio of 14.2, satisfying methodological standards against overfitting. Researchers also evaluated longitudinal modeling alternatives, but these proved less effective; a population-averaged generalized estimating equation achieved an inferior discrimination curve with an area of 0.578, while a generalized linear mixed model exhibited poor fit under diagnostic checks.
Across the testing cohort, the logistic regression model yielded an overall classification accuracy of 71.4 percent, with a sensitivity of 65.2 percent, a specificity of 83.3 percent, a positive predictive value of 88.2 percent, and a negative predictive value of 55.6 percent. Evaluation of model calibration in this independent evaluation dataset revealed a Brier score of 0.21, alongside a calibration intercept of 0.762 and a calibration slope of 0.604, with the Hosmer-Lemeshow goodness-of-fit test showing no statistically significant miscalibration. Furthermore, decision curve analysis established that applying the nomogram provided net clinical benefit across a wide decision threshold probability range spanning from 0.31 to 0.96 in the testing dataset, demonstrating its practical utility across diverse risk tolerances.
Subject of Research: Web-based nomogram for predicting the risk of subtherapeutic infliximab trough levels in pediatric Crohn's disease: personalized risk stratification and data-driven management of infliximab therapy
Article Title: Web-based nomogram for predicting the risk of subtherapeutic infliximab trough levels in pediatric Crohn's disease: personalized risk stratification and data-driven management of infliximab therapy
Article References: Yang, Q., Ge, T., Zhou, S., Li, X., Jiang, B., Wang, X., Sun, H., Zhang, T., & Li, Y. (2026). Web-based nomogram for predicting the risk of subtherapeutic infliximab trough levels in pediatric Crohn's disease: personalized risk stratification and data-driven management of infliximab therapy. Frontiers in Medicine, 13, Article 1837963. https://doi.org/10.3389/fmed.2026.1837963
Image Credits: AI Generated
DOI: 10.3389/fmed.2026.1837963
Keywords: Web-based, nomogram, predicting, risk, subtherapeutic, infliximab, trough, levels, pediatric, Crohn's, disease, personalized, stratification, data-driven, management
News Source: Harold Sullivan. (November 14, 2026). Pediatric Crohn’s Nomogram Predicts Low Infliximab Levels. Scienmag.
